Skip to content
All posts

CISA Adds Synology, LiteLLM, REDCap, Axis Flaws to KEV

November 20, 2026 · Team PlayCISO
TL;DR · the short version

On 18 November 2026 CISA added four actively exploited vulnerabilities to its Known Exploited Vulnerabilities catalog. CVE-2024-10441 in Synology DSM (fixed in November 2024) is the flaw Synology's advisory points to after it warned on 29 October of ransomware attacks on outdated DSM devices. CVE-2026-40217 is a sandbox escape in the LiteLLM AI gateway's custom-code guardrail test endpoint, fixed in 1.83.11; Google Threat Intelligence Group says an actor used it to turn limited inference keys into admin credentials in LLMjacking intrusions. CVE-2026-90817 is an unauthenticated RCE in REDCap 13.3.0 and later, exploited to deploy a cryptominer per Safe Security; fixed in 16.0.49, 17.3.10 and 17.4.4. CVE-2026-90436 is the end-of-life Apache Axis 1.x flaw exploited against Confluence Data Center. Due dates are 21 November (DSM, Axis, REDCap) and 2 December (LiteLLM).

Put it into practice
Check these CVEs in the KEV checker
See the CISA listing, due date and EPSS score for each CVE, then rank them against the rest of your backlog.
Open the KEV checker

On 18 November 2026 CISA added four actively exploited vulnerabilities to its Known Exploited Vulnerabilities (KEV) catalog: CVE-2024-10441 in Synology DiskStation Manager, CVE-2026-40217 in the BerriAI LiteLLM AI gateway, CVE-2026-90436 in Apache Axis 1.x and CVE-2026-90817 in Vanderbilt University's REDCap. The researchers and vendors who saw the attacks describe ransomware on Synology NAS devices, LLMjacking through LiteLLM, credential theft through Axis-based Confluence servers and cryptomining on REDCap servers. CISA set a due date of 21 November 2026 for three of them and 2 December 2026 for LiteLLM. Three have fixed releases; Apache Axis 1.x is end of life and will not get one.

Who this is for: CISOs and vulnerability-management leads, platform teams running LiteLLM or another AI gateway, research IT teams running REDCap, and anyone with Synology NAS devices on the network.

Based on the CISA KEV catalog and CISA's alert, the CVE records, Synology's, LiteLLM's and X41 D-Sec's advisories, Google Threat Intelligence Group's report and Safe Security's roundup, read on 20 November 2026. CISA's own entries name no attacker or reporter: each attribution below is the named source's. PlayCISO is not affiliated with any of these companies. We publish no exploit details.

Confirmed vs not yet known

CISA confirms: all four are in KEV, added 18 November 2026, based on evidence of active exploitation. Known ransomware use is "Unknown" for all four. Forensic triage is required ("Yes") for the Synology, Axis and REDCap entries and not for LiteLLM.

Vendors confirm fixes: Synology (DSM 7.2.2-72806-1 and equivalents, since November 2024), LiteLLM (1.83.11, which replaced the sandbox), and REDCap 16.0.49, 17.3.10 and 17.4.4 per the CVE record. Apache confirms there will be no Axis 1.x fix.

Reported by others, not by CISA: ransomware against outdated DSM (Synology's own notice of 29 October, which points to its advisory but names no CVE); LLMjacking through LiteLLM (Google Threat Intelligence Group); cryptomining on REDCap (Safe Security, as reported); credential theft through Axis on Confluence (Rapid7).

Not yet known: who is behind any of these attacks (no source names an actor), how many systems were hit, and whether the Synology ransomware wave used CVE-2024-10441 alone: Synology has not named the CVE, and Qihoo 360 links the same ransomware to two newer Synology flaws.

The four KEV entries

CVEProductFlaw (CISA)AddedDueFixed in
CVE-2024-10441Synology DSM / BeeStation OSImproper encoding or escaping of output; remote code execution18 Nov 202621 Nov 2026DSM 7.2.2-72806-1, 7.2.1-69057-6, 7.2-64570-4, 7.1.1-42962-7; BSM 1.1-65374
CVE-2026-40217BerriAI LiteLLM (proxy)Unprotected alternate channel; code execution via /guardrails/test_custom_code18 Nov 20262 Dec 20261.83.11 or later
CVE-2026-90436Apache Axis 1.xUnsafe reflection; unauthenticated code execution18 Nov 202621 Nov 2026No 1.x fix (end of life); see our Axis write-up
CVE-2026-90817REDCapCode injection; unauthenticated code execution with a public survey hash18 Nov 202621 Nov 202616.0.49 LTS, 17.3.10 LTS, 17.4.4 Standard

CISA's required action for each is to apply the vendor's mitigations under BOD 26-04, or stop using the product if none exist. The due dates bind US federal civilian agencies; for everyone else they are a useful clock. The BOD 26-04 calculator shows how the deadline is set.

LiteLLM CVE-2026-40217: the AI gateway flaw used for LLMjacking

LiteLLM is an open-source proxy that sits between applications and model providers such as OpenAI, Anthropic, Azure and Bedrock. Teams use it to hand out "virtual keys" with budgets and model limits while the real provider keys stay in the proxy's database. That makes a LiteLLM server a store of every provider key the organisation pays for.

The flaw. Per LiteLLM's GitHub advisory, the POST /guardrails/test_custom_code endpoint ran user-supplied Python inside a hand-rolled sandbox that could be escaped, giving arbitrary code execution in the proxy process; in the default Docker image that process runs as root. The advisory lists 1.81.8 up to 1.83.10 as affected and says the fix is in 1.83.11 (its metadata says 1.83.10), where the maintainers replaced the sandbox with RestrictedPython. X41 D-Sec, which found it, published on 8 April 2026 after an embargo, when no vendor patch was yet available.

Who can reach it. Sources differ. LiteLLM says that in default configurations exploitation needs a proxy-admin credential and rates it moderate (CVSS 4.0 5.9). The CVE record scores it CVSS 3.1 8.8 with low privileges. Google Threat Intelligence Group (GTIG) says it "observed a threat actor exploit CVE-2026-40217, a sandbox escape in LiteLLM that allows an attacker to elevate limited-access inference keys ("Virtual Keys") into full administrative credentials". Plan on the worst case: any key that can call the endpoint may be enough.

How it was used, per GTIG. GTIG's 16 November AI Threat Tracker places this exploitation in the privilege-escalation and persistence stage of LLMjacking intrusions, where "privileged service accounts were sufficient to establish persistence without the use of malware". LLMjacking is the theft of an organisation's model credentials and cloud compute to resell the access: GTIG describes "LLM mixers", underground storefronts that route buyers through pools of stolen credentials, advertised at $0.10 to $25 per million tokens. GTIG says the LLMjacking intrusions Mandiant handled in 2026 hit financial services, retail, transportation and technology organisations in North America, South America and Asia Pacific, with fraudulent charges above $100,000 in a single day and up to $1 million in the worst cases. GTIG names no actor for the LiteLLM exploitation and gives no date for it.

What to do.

  • Upgrade to 1.83.11 or later, pinned to a verified release. Do not use 1.107.0: LiteLLM says that release was published by an attacker on 21 October 2026 through its hijacked release pipeline (LiteLLM's root-cause analysis).
  • If you cannot upgrade today, block POST /guardrails/test_custom_code at your reverse proxy or API gateway, as LiteLLM advises, and keep the LiteLLM admin UI and management routes off the internet.
  • Assume keys may be exposed on any server that ran an affected version with the endpoint reachable. Rotate the master key, every virtual key and, most important, every upstream provider key stored in the proxy database. Then check provider billing and usage for spikes, new regions or unfamiliar models.
  • Look for persistence GTIG describes: new admin or service accounts, keys created outside your normal process, and new IAM roles or policies in the cloud account the proxy runs in. Practise the cloud side in the Leaked Cloud Key Lab, and order the rotation with the secret rotation order planner.

LiteLLM has had a hard year: a credential-stealing release in March (our write-up), a hijacked release pipeline in October and now a KEV entry. See BerriAI (LiteLLM) in the Breach Index.

Synology DSM CVE-2024-10441: ransomware on outdated NAS devices

The flaw. Synology-SA-24:20 (published 5 November 2024, from Pwn2Own 2024) describes an improper encoding or escaping flaw in the DSM system plugin daemon that lets remote attackers execute arbitrary code. The CVE record scores it CVSS 3.1 9.8 with no privileges needed. Fixed versions: DSM 7.2.2-72806-1, 7.2.1-69057-6, 7.2-64570-4 and 7.1.1-42962-7 or later, and BeeStation OS 1.1-65374 or later.

Useful next, if this was relevant to you

Showing 1 of 3: MCP Guard

Exploitation, as reported. On 29 October 2026 Synology said threat actors were exploiting a vulnerability it patched in December 2024 to attack devices on outdated DSM versions, aiming "to encrypt data and extort ransom", and pointed customers to Synology-SA-24:20. Its notice does not name a CVE or a ransomware family. The Hacker News reports the ransomware as C3RB3R, citing Qihoo 360 and customer reports; Qihoo 360 links C3RB3R to two newer flaws in Synology-SA-26:13. CISA's entry ties the exploitation to CVE-2024-10441 and marks ransomware use "Unknown".

What to do. Update every DSM device to a fixed build or later and turn on automatic updates, as Synology advises. Take DSM management interfaces off the internet and use a VPN for remote access. Do not stop at the 2024 builds: Synology-SA-26:13 (18 September 2026) fixes two more CVSS 9.8 DSM flaws in 7.4-90075, 7.3.2-86009-4, 7.2.2-72806-9 and 7.2.1-69057-12 or later, so the current build of your branch covers both. Check that NAS backups have an offline or immutable copy (3-2-1-1-0 checker), and look for unknown admin accounts, new scheduled tasks and encrypted shares on any device that was out of date. See Synology in the Breach Index.

REDCap CVE-2026-90817: cryptomining on research data servers

The flaw. REDCap is the electronic data capture platform built at Vanderbilt and used by universities, hospitals and research networks for studies and surveys. Per the CVE record published on 20 September 2026 by Securifera (credit Ryan Wincey), an unauthenticated attacker who knows a valid public survey hash can reach an unintended controller route from a public survey and supply a crafted file path or stream during data import, leading to code execution on the server. It affects REDCap 13.3.0 and later; the record lists 16.0.49, 17.3.10 and 17.4.4 as not affected. Public survey links are designed to be shared, so the hash is not a secret in practice. REDCap's own release notes are behind the REDCap community login, which CISA cites.

Exploitation, as reported. Safe Security's weekly CVE roundup of 23 October says observed exploitation staged a temporary web shell and then the XMRig cryptocurrency miner, disguised as image files and as the syslog-ng logging process, with persistence through scheduled tasks and an added SSH key. Safe Security does not say who observed it; The Hacker News and Security Affairs attribute the observation to Safe Security. Safe Security adds: "Patching closes the initial path but does not remove the payloads already deployed."

Separately, in June 2026 GTIG reported that a China-nexus group it tracks as UNC6508 targets REDCap servers at medical research institutions to steal credentials and research. GTIG could not confirm how that group got in and did not name this CVE, so the two should not be treated as one campaign.

What to do. Upgrade to 16.0.49 LTS, 17.3.10 LTS or 17.4.4 Standard or later. If you cannot this week, restrict or disable public surveys. Because exploitation started before the KEV listing, check every REDCap server that was exposed: unexpected PHP files in the web root, high CPU from a process named like a system service, new cron entries and unfamiliar SSH keys in authorized_keys. A server holding study data with a miner on it should be handled as an incident, with your IRB and privacy office informed. See Vanderbilt (REDCap) in the Breach Index.

Apache Axis CVE-2026-90436: credential theft through Confluence

We covered this one on 19 November: Apache Axis CVE-2026-90436 exploited: Confluence, N-central. In short, Axis 1.x is end of life and will not be fixed; Confluence Data Center is fixed in 9.2.28; N-central 2026.4 HF3 blocks the legacy APIs. Rapid7 says it detected exploitation against a Confluence Data Center server, after which the attacker collected credentials and keys and used a service account to reach a cloud data warehouse and Microsoft 365. CISA's KEV entry matches the dates in our post: added 18 November, due 21 November, forensic triage required.

Timeline

DateEventSource
5 Nov 2024Synology publishes SA-24:20 with fixed DSM builds for CVE-2024-10441Synology
8 Apr 2026X41 D-Sec publishes the LiteLLM custom-code sandbox escape (CVE-2026-40217) after embargo; no patch yetX41 D-Sec
22 Apr 2026LiteLLM 1.83.11 released on PyPI, the version its advisory names as the fixPyPI, LiteLLM advisory
20 Sep 2026CVE-2026-90817 published for REDCap 13.3.0 and laterCVE record
23 Oct 2026Safe Security describes REDCap exploitation leading to a cryptominerSafe Security
29 Oct 2026Synology warns of ransomware attacks on outdated DSM and points to SA-24:20Synology
6 Nov 2026Apache discloses Axis 1.x CVE-2026-90436Our Axis post
16 Nov 2026GTIG reports CVE-2026-40217 exploitation in LLMjacking intrusionsGTIG
18 Nov 2026CISA adds all four to KEV (due 21 Nov; LiteLLM 2 Dec)CISA alert, KEV

What security leaders should do this week

  • Search your asset inventory for all four today: Synology NAS (often bought by a department and never registered), LiteLLM (often run by a data science or platform team as a container), REDCap (usually in research IT, outside central IT) and anything bundling Axis 1.x. Check each CVE in the KEV checker.
  • Patch to the fixed versions in the table above. Rank the order with the vulnerability prioritizer: internet-facing REDCap, Synology and Confluence first.
  • Treat LiteLLM as a secrets store. Upgrade, rotate provider keys and check model-provider bills. Put a spend alert on every provider account if you do not have one; LLMjacking shows up first as cost.
  • Hunt, don't just patch. Three of the four carry CISA's forensic-triage flag, and every one was exploited before the KEV listing. Look for web shells, miners, new accounts, new scheduled tasks and new SSH keys on servers that were exposed. Walk the response in the IR Playbook Lab if your team has not practised it.
  • Close the ownership gap. All four products tend to sit outside central IT. Give each an owner in your risk register with a patch window.

None of these is a reported breach of Synology, BerriAI, Vanderbilt or Apache by these CVEs; they are product vulnerabilities. LiteLLM's separate 2026 supply chain incidents are covered in the Breach Index entry. See Synology, BerriAI (LiteLLM), Vanderbilt (REDCap) and Apache in the Breach Index.

Frequently asked questions

Which vulnerabilities did CISA add to KEV on 18 November 2026?

Four: CVE-2024-10441 in Synology DiskStation Manager (DSM), CVE-2026-40217 in BerriAI LiteLLM, CVE-2026-90436 in Apache Axis 1.x and CVE-2026-90817 in Vanderbilt University's REDCap. CISA says it has evidence of active exploitation for each. The federal due date is 21 November 2026 for the Synology, Axis and REDCap flaws and 2 December 2026 for LiteLLM.

What is CVE-2026-40217 in LiteLLM?

A sandbox escape in the LiteLLM proxy's custom-code guardrail test endpoint, /guardrails/test_custom_code. Python submitted to that endpoint ran inside a hand-rolled sandbox that could be escaped, giving code execution in the proxy process, which runs as root in the default Docker image. LiteLLM's GitHub advisory lists versions 1.81.8 up to 1.83.10 as affected and says the fix is in 1.83.11, which replaced the sandbox with RestrictedPython.

Is the LiteLLM flaw being used for LLMjacking?

Google Threat Intelligence Group says so. In its 16 November 2026 AI Threat Tracker it says it observed a threat actor exploit CVE-2026-40217 to elevate limited-access inference keys (LiteLLM "Virtual Keys") into full administrative credentials, as part of LLMjacking intrusions. LLMjacking means stealing an organisation's AI model credentials and cloud resources and reselling the access. GTIG names no actor and gives no date for this exploitation.

Does an attacker need a key to exploit the LiteLLM flaw?

Yes, some kind of authenticated access. LiteLLM's advisory says that in default configurations exploitation needs a proxy-admin credential and rates it moderate (CVSS 4.0 5.9). The CVE record scores it CVSS 3.1 8.8 with low privileges, X41 D-Sec's advisory used the master key, and GTIG says an actor used it to turn limited-access virtual keys into admin credentials. Treat any key that can reach the endpoint as enough until you have upgraded.

Is Synology DSM CVE-2024-10441 used by ransomware?

Synology said on 29 October 2026 that attackers were exploiting a vulnerability it patched in December 2024 to encrypt data on outdated DSM devices and demand a ransom, and pointed customers to advisory Synology-SA-24:20, which covers CVE-2024-10441. Synology did not name the CVE in that notice. CISA lists CVE-2024-10441 as exploited, with ransomware use marked "Unknown". The Hacker News reports the ransomware as C3RB3R, citing Qihoo 360 and customer reports.

Which Synology DSM versions fix CVE-2024-10441?

Per Synology-SA-24:20 and the CVE record: DSM 7.2.2-72806-1, 7.2.1-69057-6, 7.2-64570-4 and 7.1.1-42962-7 or later, and BeeStation OS 1.1-65374 or later. Older DSM branches are listed as "unknown" in the CVE record, so plan to move them to a supported release. Synology-SA-26:13 (September 2026) needs newer builds again, such as DSM 7.2.2-72806-9, so update to the current build of your branch.

What is REDCap CVE-2026-90817?

An unauthenticated remote code execution flaw in REDCap, the research data platform built at Vanderbilt and used by universities and hospitals. Per the CVE record, an attacker who knows a valid public survey hash can reach an unintended controller route from a public survey and abuse data import handling to run code on the server. It affects REDCap 13.3.0 and later; the record lists 16.0.49, 17.3.10 and 17.4.4 as not affected.

Who reported the REDCap cryptomining attacks?

Safe Security's weekly CVE roundup of 23 October 2026 describes observed exploitation that staged a temporary web shell and then the XMRig cryptocurrency miner, with persistence through scheduled tasks and an added SSH key. It does not say who observed it. The Hacker News and Security Affairs attribute the observation to Safe Security; a University of Florida scanner project credits a Palo Alto Networks Unit 42 report that we could not find.

Is there a fix for Apache Axis CVE-2026-90436?

Not as an Axis 1.x release: Axis 1.x is end of life. Apache published a patch and recommends moving to Axis2/Java. Products that bundle Axis ship their own fixes, for example Atlassian Confluence Data Center 9.2.28. Our separate write-up has the full detail.

Sources

Check these CVEs in the KEV checker
See the CISA listing, due date and EPSS score for each CVE, then rank them against the rest of your backlog.
Open the KEV checker
Take this with you
The four KEV entries · LiteLLM CVE-2026-40217: the AI gateway f… · Synology DSM CVE-2024-10441: ransomware… · REDCap CVE-2026-90817: cryptomining on r… · Apache Axis CVE-2026-90436: credential t… · Timeline

Ready to practise the decisions these articles describe?

Run your first Cyber War Room free →

Pro